JPH01212357A - Quantification of chromatographic information - Google Patents

Abstract

PURPOSE: To measure the sequence of peptide definitely by reducing the peptide periodically and determining the quantity of each residual amino acid thereby adapting the background level for each cycle. CONSTITUTION: A peptide sequence being analyzed for starting a method is controlled in Edman reduction cycle initially automated by a program element 11 and HPLC(high performance liquid chromatography) is performed by a program element 12 through a PH analyzer in order to identify the decomposed amino acid. Values of chromatogram are collected by a computer module and accumulated. Number of reduction cycle is incremented by a program element 13 and the process is repeated until the HPLC is performed for the entire amino acid in peptid chain. The chromatogram is identified and determined for each cycle by a program element 15 in order to determine the quantity of each PTH. Finally, background correction, delay correction and injection correction are performed by program elements 17, 19, 23.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to methods and apparatus useful in general for analyzing chromatographic information, and in particular for analyzing chromatographic information related to automated determination of peptide sequences. It is related to its use.

[Prior Art] Chemical treatments used by protein/peptide sequences were introduced in 1950 for sequential degradation of peptide chains.
It is derived from a technique devised by PehrEd in 2013 (Edian SA eta Chem).

5cand, 4,283, 1950; Edgan and Begg, Eur, J., Bioche
s, 1. 80°1967). The first step in this degradation involves the use of the E dlan reagent, phenylisothiocyanate (P).
I T C) and the amino group of the peptide are selectively coupled, and the reaction is carried out with an organic-based catalyst provided along with a coupling reagent. The second process is the cleavage of this generated amino acid from the peptide residue, and the reaction is influenced by treating the peptide with strong organic acids. Each repeated binding/fission cycle occurs on the newly formed amino group amino acid left by the previous cycle. Thus, repeated cycles effect successive separations of the amino acids that form the first structure of the peptide.

[Problems to be Solved by the Invention] Sequence processing cannot be completed by Edman reduction alone. - Once the amino acids are removed from the sample, they must be analyzed to determine its identity. The isolated amino acid derivative anilinothiazolinone (ATZ) is generally not suitable for analysis and is converted to the more stable phenylthiohydantoin (PTH) form before analysis is performed. New sequencer (W
1tta+ann -L 1ebold et al. Reference A
nal.

In this case, this conversion is carried out automatically in a reaction vessel separate from the reaction vessel in which Edman degradation occurs. The ATZ produced in each degradation cycle is extracted from the peptide with an organic solvent, transferred to a reactant, and treated with an aqueous solution of a strong organic acid to effect the conversion to PTH. The PTH produced from each degradation cycle may be transferred to a fraction collection bin until multiple PTHs are manually collected and prepared for analysis. Alternatively, PTH may be directly and automatically transferred from the sequencer conversion vessel to an online analysis system (Maehleidt, W.
and the literature by Hof'f'net, H.
Hodsin Peptide and Pr
otein 5equenceAnalysis
, pp35-47. B irr+ed,,E
l5ev1er (1980);Wlttman-
Literature Modern Methods in Protei by Llebold and Ashman
nChemistry, ppH3-327, Tsc
Hesche, ed., de Gruyter (19
85) ; Reference J by Rodriguez.

Chromotography 350, pp2
17-225. (1985)).

Although various analytical methods have been used to identify amino acids liberated during Edman degradation, only high performance liquid chromatography (HPLC) is currently in widespread use. In fact, HPLC on antiphase, silica-based backing has revolutionized peptide sequencing. It provides a rapid, sensitive and quantified analysis of PTH amino acids and is the only technique currently used for PTH analysis that can reliably resolve all of the PTH amino acids in a single chromatographic run. Furthermore, since it provides quantified data at the picomolar level, HPLC is currently the only analytical method suitable for performing microsequencing by automating an Edman sequencer.

Ideally, each chromatograph should show a single quantified answer, ie the presence of only one PTH.

In reality this can never happen.

Each chromatograph contains some of the total amount of PTH and a quantitative evaluation of the relative amounts must be performed to make sequence assignments. Multiple factors are causing this problem. First, protein or peptide samples cannot be considered pure. They always contain some level of other peptides or free amino acids that generate a PTH signal during the sequence. Repeated exposure of the sample to the acid separated during the Edman chemical reaction then
Causes splitting of the butide chain at a location other than the amino end. The newly exposed amino ends as a result of these mutual cleavages thus generate PTH after a subsequent binding/dissociation cycle. As a result, each amino acid type is
It generally changes slowly during a sequence run, exhibiting a background PTH level that typically rises throughout the first cycle and then falls during later cycles. The absolute level of background is determined by the amino acid structure of the peptide, the Edman chemistry conditions, and the molecular weight of the peptide. Third, the removal of amino-terminal amino acids in any given cycle of Edman chemistry is incomplete. Therefore, any amino acid residue that would have been removed in that cycle remains in the next binding/separation cycle and is then removed. This carryover or delay is cumulative and A variety of accidents result in a constant rate of increase in the number of molecules out of phase in the expected decomposition order.Fourth, the expected restoration of PTH is due to side reactions, which block the amino end groups during the run, from the reaction chamber. It is slowly reduced by physical loss of peptides, internal chain separations and delays. This signal reduction, measured as repetitive cycle production, occurs with increasing noise (due to the factors mentioned above) and the sequencing run Furthermore, it extends to peptides, making accurate amino acid assignment even more difficult.Fifth, the relative recovery of PTH amino acids from the Edman chemistry changes; some are mostly recovered quantitatively, while others are analyzed Almost destroyed before being destroyed.

Despite these issues, the rigorous interpretation of chromatographic data from sequencer runs in amino acid sequences does not take into account the chemical reactions and equipment used. The most common sequence is assigned by visual inspection of the chromatogram, distinguishing the specific increase in PTH levels of one amino acid during each cycle from the general background level of all PTH. This method is very simple and effective, but it has limitations. This method relies on the pattern recognition abilities of scientists and a limited technique that is highly subjective and allows only two direct comparisons of three chromatograms at a time.

Because of these limitations,1 an increasing number of scientists are using HPLC peak integration systems that convert the analog signal displayed on a chart recorder trace into a simple set of digital numbers.

This allows the recovery of each PTH in each cycle to be shown on a graph that more clearly shows the particular sequence of signals superimposed on the background noise level. S
The literature by M1thies et al. (Biochemist
ry 10.4912. (1971)) were the first to define the formula for sequencing chemistry related to initial products, repetitive products, delays and amino acid background and to attempt quantitative sequence analysis based on peak integration. Also Machioeldt, W, Chemi
stry, pp245-258. Waiter de
Gruyter.

Berlin) contributed to this process, but all previous methods have relied on the subjective classification of integrated peak values by the scientist performing the sequence analysis. The final sequence assignment still requires the scientist's subjective interpretation of the relative importance of the estimated levels of one amino acid versus another at any given cycle.

In addition, background PTH levels, cumulative delays, side reactions, etc. all of the above issues and other important issues are related to the chromatographic data itself. Although commercially available chromatographic software works well with nearly ideal data (ie, with large, well-resolved peaks), they rarely work with real-world data.

Such non-ideal data are the norm rather than the exception for the analysis of amino acid derivatives. Amino acid analysis typically involves the resolution of complex mixtures of closely related structures at extremely small levels where conventional software cannot provide satisfactory results unless the user provides extensive manual input to correct deficiencies in the software. There are many things.

Conceptually, an HPLC data system periodically
Collect chromatographic data by sampling the output of the chromatographic data and then process this digitized data. Quantification is therefore performed using peak integration, which requires locating the start and end points of the peak and measuring the summed signal between these points to subtract any background signal. The midpoint of the peak (ie, its retention time) is also required to identify it as a known component based on the retention times that are normally obtained. The measurement area of the sample peak can then be converted to a molar concentration amount based on the measurement area of the corresponding standard. However, this conceptually simple process is complicated by multiple factors, such as chromatographic noise, peak overlap, and retention time drift.

Chromatographic noise is caused by the detector electronics, incomplete mixing of the solvent during the gradient chromatograph, passage of air bubbles or particles through the detector, changes in refractive index due to solvent or temperature gradients, and solvent elution. or resulting from contamination of the column.

Current conventional HPLC systems deal poorly with both high and low frequency chromatographic noise.

Most high-frequency filtering is performed by analog filters determined by the Bird-Air instrument and installed into the detector circuit, and some HPLC systems perform point-by-point subtraction of a blank chromatogram from sample data. attempts to remove low frequency noise (often referred to as baseline drift) by using

This latter technique is particularly problematic because it introduces additional high frequency noise and the baseline can vary substantially from run to run. Clearly, new and less cumbersome techniques are required to reduce chromatographic noise.

Peak overlap, ie, incompletely resolved peaks, is particularly problematic for HPLC software.

Small peaks that partially overlap with large peaks may be missed due to the slope threshold process;
Therefore, inaccurate romatogram quantification may result.

If a cohesive peak is detected, multiple methods of dividing the total area are used, such division being done by a method in which a baseline is set below the peak component. These are (I) a linear extrapolation between the start and end points of the multiblets with a linear drop from the valley between the peaks to the baseline, and (II) a tangent that sets the baseline for the minority component. (m) a linear extrapolation between the start and end points of each separated component; The method that gives the most accurate peak measurements is determined by the degree of resolution between the peaks and the relative peak heights. It is therefore highly sample dependent and requires adjustment by the user from one sample to another in a set of chromatograms where the parameters are not constant.

Retention time drift is also problematic because once -degree peaks have been established and quantified, they must be identified by comparing their retention times to known standard retention times.

This is straightforward if the change in retention time from run to run is always smaller than the time separation between closely extracted peaks within a run. Typically, software routines are set up to search for an elution time "window" centered around the standard elution time to find the best matching known peak of the standard. The elution time drift is
This does not always work with complex separations that produce closely spaced peaks because it moves one peak outside the window and another inside. However, this problem is caused by drift.
This can be minimized by using easily identified reference peaks and modifying the search window for other peaks experimentally. The reference peak must be sufficiently separated from all neighboring peaks and present in all chromatographs such that its search window is large enough to tolerate the maximum observable drift. A method is needed to solve the problem of quantification of chromatograms, and this method is especially useful for peptide sequencing in a way that automatically arrives at unambiguous calls of the sequence without relying on the subjective interpretation of the scientist. It can be used by a computer to measure the HPLC data set obtained from the graph.

SUMMARY OF THE INVENTION According to a preferred embodiment of the invention, an apparatus and process for unambiguously determining the sequence of peptides is disclosed. Peptides sequenced according to the method are periodically degraded into a set of amino acid residues for each cycle.

The amount of each amino acid residue is quantified and measured in each set, and then the background level is adjusted for each cycle to obtain background fit. A measure of variance is then calculated relative to the background suitability, and the measured amount of amino acid residue in each cycle is normalized relative to the background suitability. The maximum prescribed background correction residual amount in each cycle thus makes a sequence allocation that can be used as needed for the next correction step if desired. Such a next step involves modifying at least some of the predetermined background correction residuals for the delays into cycles in succession, thereby obtaining a delay-corrected background correction residual for each cycle. These delayed-corrected background-corrected residues can therefore be used to correct the initial measurements of amino acid residues during each cycle for differences in injection volume between cycles, thereby injecting corrections for cycles. The previous steps of background correction and delay correction are thus performed on the injection corrected residual amount, and this maximum amount is determined for each cycle in which a clear sequence assignment for a peptide is reached.

The initial step of measuring the amount of amino acids remaining during each cycle in the preferred mode usually includes several sub-steps. In particular, as the peptides are periodically degraded using the Edman classification table, chromatographic analysis is performed for each cycle and prototypical data provides a measure of the amount of each type of amino acid found in each cycle. do. Therefore, in some embodiments, the results of the determination are low-pass filtered and bypass filtered to remove measurement noise introduced into the baseline-corrected chromatogram. This baseline corrected chromatogram is then used in the background correction process.

In another embodiment, the baseline-corrected chromatogram is obtained using a linear, motion-invariant discrete filter function aN. where N is the measured value of the filter width and α is the resolution improvement characteristic of the filter function.

In an embodiment, the chromatogram is first smoothed with a filter function that rejects high frequency noise, thereby providing a second (filtered) chromatogram having a first filtered baseline. The second chromatogram is then filtered by setting the parameter α for peak resolution enhancement to obtain a third chromatogram with a substantially similar filtered baseline. The second chromatogram is then subtracted from the third chromatogram to produce a fourth chromatogram that removes the baseline.

A fourth (baseline correction) chromatogram is then operated to generate a quantified value of PTH that is used in the background correction process.

The apparatus for carrying out the method of the invention comprises a degradation element for cyclically degrading the peptide to obtain a set of amino acid residues for each cycle, and a quantification element for measuring the amount of each amino acid residue in each set. and a computer element for controlling the degradation component and the quantification component. The computer element also fits the background level to each cycle, calculates a measure of the variance to background fit for the residual measure in each set, and calculates a measure of the variance for the background fit for the residual measure in each set, and measures the variance for the background fit to obtain a specified background corrected residual for each cycle. Standardize the measured amount of each residue and identify the maximum normalized background corrected residual amount during each cycle.

EXAMPLE According to a preferred embodiment of the invention, FIG. 1A is a flowchart illustrating a general method for determining the sequence of a peptide chain. In this method, the various steps involve both direct physical measurements and computational elements based on these measurements. The language of computer programming can be borrowed for the description of the present invention, since it can be carried out fully automatically by the measurements.Therefore, each step of the method is closely linked to that language. , referred to as either a "program element" or "subroutine 1" rather than as "step 2" in such cases. used to describe a complete method computer subprogram.The term "program element"
Freely used to describe one or more discrete steps in a set of identifiable steps forming a unique computer program or complete task, which is a physical measurement or calculation, but this in itself is not necessary. However, it does not become a signal subroutine.

The physical equipment used to carry out the method is shown in FIG. a PTH analyzer 53 that performs on-line PTH analysis, and a computer module that acts as both a system controller and analyzer for the entire system that performs the various steps used to arrive at quantified calls for the peptide sequence. 55. As a system control device, the computer module is
Responsive to time and increments of degradation results in the sequencing system to move material to the PTH analyzer and control the PTH identification process. As an analyzer, the computer module accumulates the data provided by the PTH analyzer and also eliminates noise and system errors from the raw PTH value measured by the PTH analyzer to identify the maximum relative PTH value during each cycle. Perform the various steps listed below to eliminate the peptide sequence and thereby determine the peptide sequence. In the preferred mode, computer module 55 includes a 16/8-bit microprocessor, such as an Intel 8088, a separate computational common processor, 640
It is an Applied Biosystems Model 900A that includes a kilobyte of RAM, a 510 megabyte hard disk drive, a 360 kilobyte floppy disk drive, a touch screen CRT, and a graph printer.

Also in the preferred mode the sequencing system 5
1 is an Applied Biosystem Model 470A sequencer, and the PTH analyzer 53 is an Applied Biosystem Model 120A.

The peptide sequence to be analyzed to start the method is first controlled by an automated Edman degradation cycle in program element 11 and the HPLC is controlled by an automated Edman degradation cycle in program element 12 to identify degraded amino acids.
Performed by TH analyzer. For that Edman cycle, a chromatogram is collected in digitized form by a computer module and the values of the chromatogram are stored. The number of cycles of degradation performed in program element 13 is incremented and the process is repeated until the degradation and chromatograms have been performed for all amino acids in the peptide chain, each amino acid receiving one degradation cycle. corresponds to Subroutine 15
In each chromatogram, each chromatogram is identified and quantified to determine the amount of each PTH for each cycle of degradation.

Once the degree identification and quantification is completed, a first pass in PTH background noise cancellation is performed in subroutine 17 and a preliminary sequence assignment is made. This is followed by delay modification in subroutine 19.

A delay correction is made to account for the fact that amino acid residues are only partially removed during Edman degradation so that fractional distillation of amino acids appears in subsequent cycles.

Once the -degree delay correction is done, the remaining PTH value for each cycle is
It can be used to correct for any quantitative variations in the sample injected into the analyzer. This injection correction is performed in subroutine 23. Once the injection correction is finished, the background and delay corrections are repeated and the amino acid assignment is performed in program element 25 based on the maximum corrected PTH signal in each cycle. The second pass through background modification and PTH delay modification can be completed in any number of ways.

The solution shown in FIG. 1A is that program element 2
1 to determine whether a PTH injection modification has been made, and to provide a program counter that advances by sequence selection in program element 25 if a PTH injection modification has been made. If PTH injection correction is not performed, background and delay corrections are repeated. Any of a number of equivalent program counters can be used to leave evidence of whether a PTH injection modification has been performed. For example, for the program element 21, the background correction subroutine 17, the delay correction subroutine 19 or the injection correction subroutine 23
You can use a counter for itself. Other, less effective, similar solutions not shown do not use the program counter and program loop at all, and also run background correction subroutine 17 and PTH delay correction subroutine 1 before making sequence selection.
9 after the injection modification subroutine 23.

Each of the various subroutines and program elements
This will be discussed in further detail by reference to the various figures mentioned in the Brief Description of the Drawings. Further specific details for program elements 17-25 can be found in Reference I, which presents a specific example of preferred source code implementing these program elements.

[Quantification of PTH production] Subroutine 1 to identify and quantify HPLC peaks
Flowcharts representing details of the first and second embodiments of No. 5 are shown in FIGS. 2A and 2B, and 2C and 2D, respectively. standard homepage
LC integration routines can be successfully used to identify and quantify HPLC peaks when the amount of sequenced peptides is relatively large (i.e. high HPLC signal noise level), but with modern Edman sequencers. can work by strictly testing these routines at a sample level.

The HPLC peak is (typically P from the analytical column)
both from high frequency noise (from the UV absorption detector used to monitor TH elution) and from absorption/refractive index changes (typically from temperature-induced detector drift and absorption/refractive index changes that occur during gradient-end HPLC elution). ) can be obscured by low frequency noise. Therefore, prior to peak analysis, chromatogram filtering is performed to minimize the effects of both types of noise.

According to a first embodiment, after each chromatogram is collected in digitized form by the computer module 55, the low frequency noise is detected by Goehner in the literature (Anal, Chem 50.1223 (197B)).
) is then filtered by applying the method described in ). The entire chromatogram is first fitted separately to an N-stage polynomial curve. All points above or below the calculated curve (in standard deviation units) above or below the selected total are therefore removed and a new fitness is generated using the remaining points. This process continues until the standard deviation between the actual and calculated points reaches a set minimum level (i.e. the adapted curve is focused on the baseline of the chromatogram). Ru. The polynomial coefficients of the calculated curve can thus be used to calculate the baseline point for each point in the initial chromatogram. That is, 2
The difference between the set of values represents the baseline corrected chromatogram.

In fact, the variation in the number of initial chromatograms and the number of slopes of the baseline (this variation determines the number of steps N of the polynomial used for adaptation) - exceeds the limits of the practical computational power of microcomputers for boat fishing. It is something. The routine is therefore run sequentially on a segment of the chromatogram that is common to each round of the fitting routine used to provide a background for a portion of the total chromatogram. Typically the first 3/9 of the chromatogram points are fitted first to fit the first 5/9 of the baseline points.
18. Then point 3 starts 2/9 of the way from the front edge of the chromatogram.
79 are fitted and used to set the next 4/18 of the baseline points.

The 3/9 points starting 4/9 from the front of the chromatogram are then matched and used to set the next 4/18 of the baseline points.

Finally, the last 3/9 points are fitted and used to set the last 5718 baseline points. Irregularities in the baseline-corrected chromatogram where the individual fitted curves intersect are minimized, since only the central part of each calculated curve is used when possible.

The baseline calculation process of this first embodiment is shown in detail in FIGS. 2A and 2B. The process begins with program element 211 where all program counters and constants have been initialized. In particular, at least "i", 1k", "m".

Three counters are used. Counter i is used to index the unique part of the chromatogram that is being fitted, the counter is needed to identify the unique Edman cycle that is being fitted, and counter m keeps track of the number of iterations of the fit. Must be used to save. It is also necessary to select the N-stage polynomial used for fitting the chromatogram and determine the desired approximation of the fit to the background calculations. In practice, N is typically chosen to be about 6, and an approximation to the fit is generally such that the standard deviation of the fitted curve to the background points is a constant K (the high frequency noise level of the detector output). (typically twice). - Once the variable constants and counters have been initialized, the method continues in program component 213 by fitting the i-th part of the measured chromatogram C(t) corresponding to the 1st Edman cycle. On the bus of t- = 1 and the N-stage polynomial P\ (t, N) (i.e., the door f, (t, N) in the first iteration) typically uses a least squares approach. C8 (t). In program element 215 the standard deviation σ1.(c) is calculated for the conformance and also corresponds to the maximum point-to-point deviation allowed. Function δ(σ
, , ) are determined. Here, δ (σth) is 0
, lσ1la. The magnitude of the standard deviation σ-6 is tested in program component 217 to see if it is less than or equal to a selected constant. If it is less than or equal to that, counter i is incremented in program element 223. Counter i is then tested at program element 225 to see if it is greater than four. If it is not greater than 4, the next part of the chromatogram for the kth cycle is
Compatible with

If the standard deviation magnitude is greater than K in program element 217, the program increments the iteration counter by m in program element 219.

Then in program component 221 (this first bus C
``s' (in t)) new function Ck,' (
t) is calculated, and then all points j from the chromatogram are calculated as l C−r (t) −p, (t, N)I >
By removing it as S (σffi ), it is called a reduced chromatogram. This reduced chromatogram has the same values as the initially allJ determined chromatogram, but its domains have been reduced. Therefore, starting this reduction chromatogram again in program 213 passes through program elements 213 to 225 until the reduction chromatogram for all parts of the chromatogram of the cycle 213 meets the set compliance criteria. It is compatible with new polynomials etc.

Once the polynomial fitting has been completed for each portion i of the chromatogram of the kth cycle, the baseline of the kth cycle is calculated in program element 227 and the value of the fitting polynomial phi' (t) is determined. used to calculate the baseline point b' (t) for each point in the chromatogram Co (t), where 1=max, m
(final iteration). Then the background corrected chromatogram Co (t) is calculated from the baseline point b' (t) of the measured chromatogram Go
(t) by subtracting from program element 22
9, which corresponds to removing low frequency background components from the chromatogram. Those skilled in the art will understand that the above solution for filtering low frequency noise in this embodiment is only one of many similar solutions. For example, any complete set of functions can be used for functions that fit better than polynomials.

Once these low frequency components have been separated from the chromatogram, high frequency noise is removed in program element 231 using standard fast Fourier transform filters known in the art. Peaks in the filtered chromatogram are therefore detected and standardized in program element 232. Multiple methods can be used.

For example, a time window established based on the observed elution time of each PTH in a standard mixture can be applied to a filtered chromatogram C(t) to determine the amount of each PTH represented in the chromatogram. can be used to search for.

Quantities can be determined by finding a standard first-occurring peak and a known peak integration process, or by fitting the points in each window to a Gaussian curve and calculating the area under the curve (as described by Kent et al. Literature B 1 ot
echnlques 5. pp314 to 321 (19
78)) can be measured. Or even more simply, the maximum black value in each window corresponds to PT
It can be thought of as the height of the H peak.

Although this solution is quick, the PT
Accurate separation of H is necessary.

Performing the peak location and quantification results in a transformed chromatogram PTH(k) during each cycle, which corresponds to the amount of PTH of the j amino acids in the cycle. This program is then tested for cycle numbers in program element 233 and if all of the cycles requested to degrade the peptides in the sample have not been completed, the cycle number is
35 and the background modification process begins again at program element 213 for a new cycle.

Figures 3A(1) and 3A(2) show the results of using the technique described in Figures 2A and 2B for this first embodiment to remove low frequency noise. In (1) of Figure 3A, the uncorrected chromatogram Ca
(t) is 5 pa+ol due to its many peaks
Shown as a function of time for the P T H standard. The fitted baseline b' (t) is the curve cg
It is shown as a relatively smooth dark solid line at the bottom of (t). FIG. 3B(2) shows the baseline corrected chromatogram of the sample as determined in accordance with the method of the invention at C8(t), program element 229.

As another preferred embodiment, another approach can be used to quantify the chromatogram using digital filter techniques.

Although it is known that digital filters generally smooth noise spectra and increase resolution, and have been applied specifically to physical measurements obtained by ENDOR (Electron Nuclear Double Resonance), they have clearly not been applied to chromatogram quantification. Ta.

(For a general discussion of digital filters in noise reduction, “Variable Filters for Digital Smoothing and Analysis Enhancement of the Noise Spectrum”, Broo+ba et al., Anal, Chem, (1984), 56
．． 2052-2058, and “Characteristics of Variable Digital Filters for Smoothing and Increased Resolution”, B
lermann et al., Anal, Chew.

(1986), 58, 536-539. ) The particular approach disclosed in these references involves the use of a digital filter that is a discrete, linear, transform constant convolutional motion defined by the following equation:

where A represents the filter operation, a (n) is the filter function (kernel), f is the unfiltered spectrum, and Af is the filtered spectrum. In particular, the filter function is a vertically shifted triangular filter as follows.

(2α+1)2α1nl a (n) = 2N+I N (N+1)(
n<N), this filter function is generally divided into two parts as follows.

a (n) = al (n) 10a2 (n) where, 2α a 1 (n) = (N+1- I n l) N
(N+1) triangular filter, and 2α(N+1) −N C2(n)” N (2N+1), which is a rectangular filter. From now on, this combination is defined as the filter function al(
n) and C2(n) have the appearance of a house, “
It is called “house” filter. α-1
, N-9 for an example of a triangular filter superimposed on a rectangular filter using N-9.

Figure 3B (2) shows a synthetic house filter. ) In this filter function, N is the filter width and α
determines the degree of resolution increase. The basis for the increase in resolution with this filter is the reduction in line width, since the frequency response exceeds 1 for α>1 at low frequencies. At higher frequencies, this frequency characteristic is sharply reduced and high frequency noise in the signal is suppressed. The characteristics of the filter naturally vary depending on α and N. In general, α-1/
2 is particularly suitable for signal-to-noise enhancement of unknown spectrometric functions, close to matched filters,
It also produces the most appropriate high frequency attenuation.

In general, α-1 is the maximum value that prevents the frequency response from exceeding 1, so α-1 marks the boundary between the smoothing filter and the resolution increasing filter. For α>1 the frequency characteristic of the filter increases above frequency f-0 and then drops sharply giving a general resolution increase.

For α〉〉1, the resolution increase increases monotonically with α, but amplifies the noise.

Although these digital filter techniques could be directly applied to the chromatogram quantification problem, due to the systematic errors inherent in chromatographic techniques at n1 constant times, this result still contains residual low frequency noise. . To avoid that problem, another approach is used to construct the second embodiment of the invention. This second approach is illustrated in the flowcharts of FIGS. 2C and 2D.

As far as the previous embodiment is concerned, the computation begins by initializing constants and calculating this time in program element 238. In particular, a counter corresponding to the number of Edman cycles is set to 1 so that calculations are processed cycle by cycle. At program element 240, the filter parameter is generally set equal to a number between 1/2 and 1, and in the preferred embodiment is set equal to 1. The filter width is set to match the chromatogram peak width N1, so in the preferred mode this filter function is similar to the Savitsky-Golay
zky-G olay) filter, and supports smoothing without increasing resolution. Program element 242
, the chromatogram measured for cycle k, ie Co (t), is filtered by convolving it with a house filter kernel. This produces a smoothed chromatogram Ct'1(t), ie high frequency noise has been filtered out. α is then set equal to a number greater than or equal to 1 for increased resolution in program element 244, in which case it is set equal to 20 in the preferred mode. Program element 24
6, the smoothed chromatogram Ct'+ (t
) is the increased smoothed chromatogram CI2
(t) is rotated by the new house kernel to yield (t). This chromatogram typically has essentially the same baseline as the smoothed chromatogram C (1) but with increased peaks and negative sidelobes because the house kernel protects the area. have Consequently, by subtracting the smoothed chromatogram C+'r (t) from the smoothed chromatogram CI2 (t) incremented in program element 248, a chromatogram retaining only peak and sidelobe data is generated. We obtain the gram CI'3 (t).

Furthermore, low frequency baseline noise is completely removed, thereby eliminating an important source of error in chromatogram quantification. This peak is important information, so
The sidelobes are the new chromatogram CI'4 (t
) is truncated at program element 250 to generate. Typically this is done by removing those points below a given threshold (generally zero). At program element 252, the conditions for the new house kernel are to set up a matched filter for the chromatogram C(t) by setting α equal to 1/2 and width N equal to N1/2. established by. A new chromatogram, Cps (t), is then calculated using this new kernel in program element 254;
It is essentially a truncated chromatogram CI'4
(t) is smoothed.

In program element 256, chromatogram C115(1
) is detected and quantified, and program element 258
, these peaks correspond to specific E with specific amino acids.
In order to correlate each of the peaks in the dman cycle, the quantified values PTHj,o are compared with known reference peaks for different amino acids.
(k) is obtained. At program element 260, the cycle number is tested to see if it is greater than or equal to the number of amino acids in the peptide. If so, the program continues to subroutine 17. Otherwise, the cycle number is incremented at program element 261 and chromatogram quantification is restarted at program element 240 for a new cycle. The reference peaks used for the above comparisons are run through a chromatogram on pure samples of each amino acid and a filter routine is applied for each of them to obtain a filtered reference chromatogram for comparison with the sample chromatogram. Obtained by running.

Figures 3C-3H show the results of applying the smoothing and baseline subtraction method described above to a standard PTH mixture having the uncorrected chromatogram of (1) in Figure 3A. Figure 3C shows the unmodified chromatogram of Figure 3A (1) on a reduced scale. The results of filtering the chromatogram of FIG. 3C using a house filter with N=3, α-1 to remove the high frequency noise of FIG. 3D is shown. Figure 3E again shows the result of applying the house filter (ie, applying it to the filtered chromatogram of Figure 3P), this time at N-3, α-20. Third
Figure F shows the result of subtracting the chromatogram in Figure 3D from the chromatogram in Figure 3E. Figure 3G shows the chromatogram of Figure 3F after cutting at zero. Figure 3H is N
Figure 3G shows the chromatogram of Figure 3G after smoothing with a house filter with α-2, and α-1/2. Third
■The figure is shown in order to obtain a peak height that is more comparable to the peak height in Figure 3C in order to compare the two curves.
Figure 3H shows the result of scaling the chromatogram of Figure 3H (divided by an arbitrary scale factor of 3). To illustrate the sensitivity of this method, these results are shown in Figures 3C' and 3.
Figure D' and 3rd! ' are shown using enlarged proportions in Figures 3C, 3D, and 3 respectively.
■Correspond to the diagram.

To further illustrate the sensitivity of this method using a house filter, the results of application of this method to the 20 pffloI PTH standard are shown in FIGS. 3J-3M. 31st
The figure shows the uncorrected chromatogram for the sample.

Third, the figure shows the uncorrected chromatogram at an enlarged time ratio in the time interval between 14.0 and 16.6 minutes, which clearly shows the substantial high frequency noise in the measured chromatogram. Indicates inclusion. Figure 3L is N-6,
The results are shown using a house filter to remove high frequency noise having α-1. Figure 3M shows the result of applying the all-digital filtering method to the unmodified chromatogram of Figure 31, showing that the final filtered chromatogram is completely smooth, with peaks that are completely successfully removed. show. The next filter parameters used for resolution increase were N-6, -20 and for final smoothing N-3, -172.

PTG Background Correction Processing of all chromatograms from a sequence run by the routines described above in Figures 2A and 2B or Figures 2C and 2D will
Produces a data array containing unmodified values for TH. Plots of individual PTH versus cycles typically show rising and/or falling background levels of PTH with one or more cycles such that the PTH value is substantially above this background level. This background level is removed from the remaining PTH production by a circular minimum square variation suitable for the polynomial routine used earlier for low frequency filtering of the chromatogram. In this variation, the iterations of the adaptive algorithm are continued until the ratio of standard deviations between the actual and fitted data points for the sequence iteration is greater than or equal to the set value, ie S. When this iteration is finished,
The calculated background PTH value is subtracted from the uncorrected value to yield a background corrected PTH value.

An estimate of the variance of this background corrected data is then made by running three iterations of a least squares polynomial fitting routine (of the order of polynomials minus 1).

The standard deviation of the final iteration then provides an estimate of the change in background level, an estimate that allows the high level in a particular cycle to actually be increased to a statistically relevant amount. Once the iteration is finished, the calculated background corrected PTH value is divided by the standard deviation of the background suitable to obtain a normalized background corrected PTH value. This process then ensures that the remaining PTH value is greater than or equal to the value calculated from the fitted background curve (
or less) is repeated for each PTH, also expressed in units of standard deviation. In this regard, a preliminary sequence assignment is made for each cycle by selecting the PTH whose background corrected normalized value is statistically highest in that cycle.

This process is explained in further detail in the flowcharts of FIGS. 4A and 4B. The process now begins at program element 411, where required constants and program counters are initialized.

Program counter "j" is initialized to the value 1.

Counter j is used to display the PTH value because it indicates j types of amino acids. The integer M is set equal to the number of cycles divided by 12 (expressed to the nearest integer), which corresponds to the rank of the polynomial used to fit the PTH value. Also array δ1. -6 is set equal to zero for all s-1 through 20. In addition,
The value of S is set equal to 2, and S is the cutoff value of the ratio of standard deviations between the actual and fitted data for successive iterations to determine when the iterations are terminated in the least squares fitting routine. It is. Program element 41
At 5, another counter "1m" is initialized to zero,
This counter 1 corresponds to the number of iterations during the least squares fit routine. In program element 417, PTH amount, *
(k) is an M-class polynomial Q +, s+r (M,
K). First this means PTH1,o (
k) (the sum of values for amino acid number 1 as a function of cycles) is fitted with the polynomial Q+, r (M,k). At program element 419, the standard deviation of the fit to the jth amino acid on the first iteration is calculated and δ1. s
(i.e. initially δ1+O), and the fitness function Δ
A measured value of (δ1.1) is calculated. Here Δ(δ4.
．． ) is the standard deviation δj for points higher than the uncorrected data
, -, and twice the standard deviation for points below the uncorrected data.

The ratio of δ,1-1 to δ,1 is then calculated in program element 421, and if the ratio is greater than S,
Normalized P background corrected with program element 425
Proceed to calculate the TH amount. If this ratio is less than or equal to S, counter 1 is executed in program element 422 and in program element 423 a new PTHl, * (
k) is the amount of IPTH, r-+ (K)-Q+, s
(k) is calculated by moving those points (i.e. cycles) such that 1>Δ (δ11 ). PT for all points here
H5,s (k) - the PTH quantity, n (k), is common to the domain of both functions. When this point is moved, the fitting routine is restarted at program element 417 and the iterative process is performed with ratios δ, 1/δ31 . This continues until S becomes larger than S. An iteration counter m is then initialized at 424 and, as shown above, at program element 425, the background corrected uncorrected PTH amount, PTH amount 5. s (k
) is calculated for each cycle. This gives the final fitted value Q+,a for PTH of amino acid j in the cycle from the measured PTHffiPTH+,*(k)
This is done by subtracting (k). In the program element 429, the background corrected uncorrected PTH amount PTH'
r,s (k) is a first degree polynomial Zr, ,+r
(k) (Z+, + (k) is the first iteration). Compatibility δ′19. The standard deviation of and the measurement of the fitness function Δ′ (δ′+, m)
It is calculated by 31. where Δ′ (δ′ 11.) is generally chosen to be equal to the standard deviation for points falling on the fitted line and equal to 1.67 times the standard deviation for points falling below the fitted line. be done.

This program then uses program element 433 to stop repeating at the third time for this particular sequence of adaptations.
tests the value of the iteration counter m. The iteration counter is then incremented in program element 435 and the new PTH amount is incremented in program element 437. , (k) is the amount of PTH. (k) and the fitted value is calculated by moving those points from the background corrected uncorrected value such that the fitted value is greater than the measured value of the fitted function. This iterative process is then repeated two more times, for a total of three times, after which the normalized and background corrected PTHtlP
T H... (k) is calculated in program element 439 by dividing the background corrected uncorrected value by the standard deviation calculated in the final iteration.

Background corrected normalized PTHffi is amino acid j
counter j is calculated for program element 441
and j is incremented by program element 44.
2 will be tested to see if it is less than 20. If it is equal to or less than 20, the program loops back through the background adaptation and modification process, and the program element 415
resume by. If j is greater than 20, the method proceeds to program element 443 where a preliminary sequence assignment is made by finding the maximum scaled amount of PTH for each cycle.

Those skilled in the art will understand that there are many ways to perform background correction. For example, different fitting functions may be used and different measures of standard deviation and variance may be used for fitting. An important aspect of this background correction process is to arrive at a standardized sequence of PTH quantities so that results from cycle to cycle can be quantitatively compared.

Delay Correction After the scaled background correction PTH9 has been computed and the preliminary sequence assignments have been made, delay correction is performed. E
In the cycle of da+an degradation, the removal of amino acid residues is partial, ie a small amount of amino acid appears at +1, k+2, . . . , k+i in the subsequent cycle. In any given cycle, a failure to combine and/or split typically results in 1 to 2 out-of-phase signals or delays.
Add %. As these failures accumulate, the observed delay becomes progressively larger as the sequence progresses,
Cycle n+ from cycle n during long-term driving
1 more signals appear. For this reason, delay correction is particularly important for accurate sequence assignment for subsequent cycles.

If “YO” is the theoretical initial signal and “b” is the failed part of the reaction, amino acid 1 appears during cycle j.
If we define "Yl," as the amount of produced , Yr, s -IYo (1-b) Yl, 2 = IYO (1-b) b Yl 13- I Yo (1b) b2yl, 4
-IY, (1-b) b3 and 6° Similarly, for amino acids 2 and 3, Y2.2 = IYO (1-b
) 2 Y2 , 3 =2Yo (1-b) 2bY2.4-
3YO (1b) 2b2 Y2 , s -4YO (1-b) 2b3Y3.3-I
YO゛(1-b) 3 Y3 , 4-3Yo (1-b) 3 bY3 , s −
6YO (1-b) 3b2Y3.6-10Y, (1-b
)3 b3 This delay occurrence is expressed as a ratio to the observed first cycle occurrence, and these expressions are reduced to:

Yl, 2/Yl, 1 -1b"l +
3 /Yl, 1 -1 b2Y1 +
4/Y1, 1-11) 3Y2,
3/Y2, 2-2b Y2・4/Y2・=3 b2Y2,
s/Y2, 2-4b3y3. 4
/Y3 , 3 ”3bY3 ・ 5 /Y3 ・
3.5b2Y3, 6/Y3, 3-
10b3-In terms of boat fishing, if Y, is the first cycle occurrence (i.e., Y, , ), and Y k+1 is the delay occurrence (
That is, if Yk is +1), then Yk+l
/Y, ~C[k+O] /1) bYk+2/Yk - ([k+11 /2) ([k+0] /1) b
2Y k+s / Y − −([k+2]/3)([k+1]/2)([k+O
] b3...Y*++/Yk - ([k+1-11/i)... ([k+2]/3) ([k+1]/2) ([k÷0]
c) bl This expression is not strictly modified because it assumes that there is no irreversible signal loss and that the failure portion is the same in each cycle. However, the former assumption leads to relatively small errors as long as the irreversible losses are less than 10% per cycle and are therefore not significantly disturbed by subsequent calculations. The effect of the latter assumption, which is clearly inaccurate, is more difficult to assess. Empirically, b does not appear to interfere when measured each cycle as a cumulative average delay.

A preferred method of delay correction is illustrated in FIGS. 5A, 5B, and 5C. Program element 511
In , the preliminary sequence assignment determined from subroutine 17 is assigned to the first cycle generation array Yh (i.e., Y,
−PTH, (maximum of (k)).

This preliminary sequence assignment is then used to calculate the operating value for the cumulative delay, kb, at each cycle.

This calculation is performed by program elements 513, 515, and 51
Commenced at 7. First, program element 513
So, Yh, h++ are all E da+an in the sample
is set equal to P T Hr (k + 1 ) for the cycle, where j is selected corresponding to the amino acid selected in the preliminary sequence assignment for the cycle. Cycle counter "n" is then initialized in program element 515 so that each cycle is modified simultaneously, and the operating value for delay modification is determined by the equation Y
*, ***/Y h −k b(k) for the remaining cycles (i.e., here k
in). In program element 519, these operating values of kb(k) are then fitted to a polynomial curve B(k) of rank Z-N/15 ("to the nearest integer"), using the method of least squares. Ru.

Measured values of delay coefficients that differ from the fitted value by one or more standard deviations are then discarded and the remaining data points are refitted. To achieve this, all E dl
The actual measured value Y t, − on the region N of the lan cycle
The standard deviation δ, of the fitness function B(k) from +t/Y* is calculated in program element 521. In program element 523, all points k for which the actual value Y *, *** / Y k differs from B (k), i.e. from the fitted value, by a standard deviation δ greater than or equal to 1 are moved from the area N and a new area N is created. ′ is formed. A least squares polynomial fit of kb (k) is then performed in program element 525 using the new domain. In program element 527, a corrected adapted delay value B' (k) is generated for every cycle from -1 to N.

The failed part b (n) is then calculated in program 529 using the fitted value of B' (k) for cycle n, and in program element 531 b (n) is calculated using the equation, G (n, t, b ) = ([n+4-11/i
)...([n+2] /3)([n+11 /2)(
n+o]/1) to the next few cycles using b' (N) or G (n
, t, b) is applied to every cycle i until the delay ff1G (n, i, b) < 0.01, i.e. until the cycle n yield is less than or equal to 1% of the cycle n yield.
used to generate.

In program element 533, the normalized PTH values are divided into three maximum values Yζ, Yi '1 and Y-! and their corresponding delays Y n l 1141 s Y#
*, l++ 1 and Y I + all are separated for cycle n. At program element 535, the maximum value is tested to see if it is less than or equal to the next highest value by a factor of 3, and if it is not (i.e., it is greater than or equal to 3 times the next highest value), and this program It is program element 537 such as Y[, etc.
Also by setting , leave the amino acid allocation when . However, if Yt is less than or equal to three times Ya, then the three maximum values Y[, Yt, Y;'
uses the adapted delay based on the original sequence assignment to determine whether the delay modification makes any change in the sequence assignment.
The delay is corrected at 7. In particular, the adaptation delay Y1'+'
Each of G (n.

i, b) calculated as Y4, if the actual Yjll
l. If - is above zero, the occurrence Y7 is modified due to the delay in cycle n+i. This modification is in program element 547, if μJEj ” 7 + nil > Y Il+1 then Y
By adding Y7 to n + l or yfill > Yi. . If 1, go to Y6 Y51.
. This is done by adding 1. This modification process continues for the next few cycles i past cycle number n, and the criterion for the cutoff is that the adaptation delay is less than or equal to 1% of the activation value as tested in program element 549. It is. This process continues for each cycle i and each Y! until each Yi' is modified by replacing with and Y(.

At program element 553, each of these delay corrections is tested to find the maximum delay correction value, the sequence value Yn is set equal to its maximum, and the amino acid index selects the appropriate sequence request for that cycle n. It is determined because When a sequence request for a period of cycle n is terminated, after either program element 537 or program element 553, the conformance delay for the period of that sequence request is calculated in program element 559 and the next number of cycles past cycle n are calculated. During that cycle, the same termination criterion is used again until the adaptation delay is less than 1% of the actual delay. Cycle n is then determined by a subsequent cycle having a positive delay value using the same cutoff criterion as before by increasing cycle n occurrence to the lesser of the calculated delay amount in program element 561 or the actual cycle occurrence. Corrected for delay from cycle n+i. The next cycle n+i is determined during all cycles i for which the adaptive delay factor is less than or equal to 1%, between zero or the observed occurrence Y7°, +1 in program element 563 and the calculated delay Y1+2. is corrected for the delay value from cycle n by replacing the delay value at cycle n+i by the greater of the differences.

At this point, cycle n is adapted and cycle n+1 disturbs the sequence assignment of cycle n+1.
There is no delay from At program element 565, the amino acid assignment is modified based on the calculated delay correction, and at program element 567, a cycle counter n is incremented to calculate the delay during the next cycle. This cycle counter is tested in program element 569 to see if all cycles have been modified. If they have not been modified, the method returns to program element 517 to determine an empirical delay calculation. The polynomial fitting routines 519-525 are repeated based on the new sequence assignments and delay coefficients, and the process continues until the delay during cycle n+1 is corrected and the delay from cycle n+l is removed from the next few cycles. . This procedure continues until all cycles are corrected for delay simultaneously. When delay corrections are made during each pass through the procedure, one or more cycles will not delay the amino acid assignment until eventually all cycles have been assigned without regard to delay effects. The process then proceeds to the installation modification subroutine 23 or is output depending directly on whether the result of the sequence assignment or installation modification has been made. Those skilled in the art will appreciate that there are other approximations to which the effect of delay modification can be arrived at. For example, instead of using the three highest values of normalized PTH, the highest values can be used, or the two highest values can be used to see if the sequence assignment changes; If you do that, go back to the other sequence selections and check. Similarly, if it is clear that the sequence request is ambiguous after delay correction, choose the highest value of three or more.

When the introduction correction background and delay corrections are made, the remaining PTH values at each cycle are used to correct for any changes in the amount of sample introduced onto the PTH analyzer at each cycle, if desired. During each cycle,
All but the two highest PTH values are averaged. When the corrected PTH value is within a standard deviation unit, this average is close to zero if the introduction for any given cycle was accurate. Any non-zero mean during one cycle is calculated by subtracting the product of the corrected cycle mean and the standard deviation unit of that PTH (as calculated in the PTH background correction routine) from each uncorrected PTH value. This is used to correct the uncorrected PTH generation data. This procedure effectively uses a set of unassigned amino acid values, such as internal sampling standards, at each cycle. Therefore, the introductory modification can be applied to any added internal H
This is done when there is no PLC standard.

A flowchart illustrating the installation modification is shown in FIG. First, in program element 611, array t
z+, a) is defined, here element Z11. ．． is calculated from the delay correction subroutine 19,
represents the value of PTHffi of j types of amino acids in lan cycle n. In program element 613, this array contains the two maximum values during each cycle, namely Zl', 11 and 2.・
-9, classified by amino acids to determine. This array is then averaged in program element 615 for those amino acids other than the two highest values, (7n)
yields a column array defined as . Introduction modified P
The TH value is then calculated in program element 617 by equation 1, where PTH, . (n) is the uncorrected PTH value found in subroutine 17 on final iteration 1 for amino acid j. Finally program element 61
At 9 the array PTHI,o(n) is set equal to INJ+= to construct the required name of the PTH array used in subroutine 17.

゛Once all of the introductory corrections to the uncorrected PTH data settings have been made, the PTH background and delay corrections must be calculated again using this adjusted data. 1゛Once this has been done, the following amino acid The assignment is error free as well as possible given the starting chromatogram.

Utility of the Invention Appendix ■ is a series of tables illustrating the results of the series of modifications described above during several steps of this method. Table 1 is 1
Uncorrected data resulting from 60 cycle quantification of an 8 kilodalton chain (ie via program element 15) of the E dllan sequence is shown.

Table 2 shows the same uncorrected data as the first row, ie, aspartate, with background correction according to the present invention. Table 3 shows the results of the next loop through the background correction subroutine 17 to correct for the background in the second row, ie, asparagine. As previously explained, this background correction process is repeated until the amount of PTH for each amino acid is background corrected.

Preliminary sequence assignments can then be made. Table 4 shows the results of background correction. The rotated elements of Table 4 are the maximum PTH values at each cycle and correspond to the first preliminary sequence distribution.

Using the preliminary sequence assignment, the delay modification subroutine 19 is then executed followed by the introduction modification subroutine 23. Once the installation modification has been completed, the installation modified sequence of PTH is then background modified in program element 17. The results of the delay correction on the background corrected data during cycles 1-37 are shown in Table 5. For comparison, Table 6 shows the results of delay correction on background corrected data during cycles 1-38. Table 7 shows the results of delay correction during all 60 cycles. The results of the introductory modifications during cycle 1 are shown in Table 8. Table 9 shows the results of the introductory modifications between cycles 1 and 2. The installation correction process then continues for each cycle until all cycles have been delayed corrected. The results are shown in Table 1O.

Table 11 shows the results of the background correction on the introduced corrected data of Figure 1O. FIG. 12 shows the result of delay correction during a full cycle of the background corrected data of FIG. Again, the maximum value in each cycle corresponds to a sequence assignment, which was found to be different from the preliminary sequence assignment shown in Table 4, and an introductory modification was performed.

This difference in sequence assignments is shown in Figures 7A, 7B, and 7C, which illustrate the results of the steps in the above sequence during cycle 51 of 18 kilodalton protein.
It can be seen more clearly in the figure. In FIG. 7A,
The reserve allocation is revealed to be lysine during that cycle. When the introduction modification is made and the background is modified, cycle 51, as shown in FIG. 7B.
Although the choice for is less clear, it appears to be histidine. This selection is then confirmed in performing delay correction as seen in Figure 7C. The effect of correction processing is 8th A.
Also seen for specific amino acids, as illustrated in Figures 8B and 8C, is the result of a corrective process for glutamate through cycling. 8th
Figure A shows uncorrected data (introduction corrected) for glutamate by cycle. Figure 8B shows the results after background correction of the data in Figure 8A, and Figure 8c shows the results after both background and delay correction.

Those skilled in the art will appreciate that there are many similar ways of carrying out the above method and that different combinations of equipment may be used to accomplish the method. For example, computational program elements may be configured separately from a computer module so long as the equipment used for their calculations is under appropriate control of the computer module. In addition, the particular program counters and constants selected in the preferred embodiment may vary depending on, for example, the number of cycles decremented and the desired accuracy of the calculations and the desired accuracy for terminating the peptide sequencing request. changes over time. For these and other reasons, the scope of the invention should be construed with reference to the appended claims and not limited to the particular embodiments chosen to illustrate it. .

Figure 3A shows the original chromatogram of a standard PTH mixture at (1); Figure 3B shows the two components of the house filter according to the method of the invention; Figure 3B shows the two components of the house filter according to the method of the invention; Figure (2) shows the house filter performed by adding the two components of Figure 3B (1), Figures 3C, 3C', 3D, 3D' and 3E. FIGS. 31 to 31' are 3A (1
) shows the results of applying the filtering method according to the present invention to the standard PTH mixture shown in the figure, and the third C'
Figures 3D' and 31' are similar to Figures 3C and 3D.
31 to 3M show the results of applying the filtering method according to the invention to a second standard PTH mixture; FIG. 4A Figures 4B and 4B are flowcharts illustrating the method used for background level correction in the amount of each amino acid residue for each cycle; 6 is a flowchart of a method used to correct for delays in injection; FIG. 6 is a flowchart of a method used to correct for differences in injection volume for different cycles; FIGS. Figure 7C shows the results of applying the filtering method according to the invention at different stages of the modification process to the 51st cycle of protein degradation;
Figure A shows the injection-corrected prototype data, Figure 7B is the result of the method after performing background corrections on the injection-corrected prototype data of Figure 7A, and Figure 7C shows the background corrections shown in Figure 7B. 8A to 8C show the results of the method of the invention at different stages of the correction process for different cycles of glutamate in protein classification; Figure 8B shows the result of the method after performing background correction on the injection-corrected prototype data of Figure 8A, and Figure 8C shows the result of the method shown in Figure 8B. This shows the result of performing delayed correction after background correction.

Applicant's representative Patent attorney Takehiko SuzueFig, IB 9.00 15.0 21.0
27.0.010 ALIFS,
#Fig, 3C (1) 9.00 15.0 21.0 27.0
．． 002 AUFS #Fig, 3
C(2) 9.00 15.0 21.0 2
7.0.010AUFS# Fig, 3D (1) 9.00 15-0 21.0 2
7.0.002 AUFS #Fig
, 3D (2) 9.00 15.0 21.0 2
7-0.010 AUFS, t9.
00 15.0 21.0 27.
0.010 AUFS #9.00
15.0 21.0 27.0.
010 ALIFS #Fig,
3G 9.00 15.0 21.0 2
7.0.010 AUFS #Fig
, 3H 9.00 15.0 21.0 27
．． 0.010 AUFS 'Yu, 002 AUFS# 1, Indication of case Patent Application No. 1983-149912 2, Name of invention Quantification of chromatographic information 3, Person making amendment Relationship with case Name of patent applicant Upride Bio Systems Incorporated 4, Agent address: 3-7-2 Kasumigaseki, Chiyoda-ku, Tokyo Specification (brief description of drawings) Drawing 7 Contents of amendment (1) Specification page 6s No. 18 In line 1, "Figure 1A is" is corrected to "Figure 1A is". (2) On the same page of the specification, line 20, it is written as "Rat diagram B is". (3) In the first line of page 70 of the specification.

``Figure 2 A and 2 B are Φ...''
The figure and FIG. 2B are corrected as "...".

(On the 3rd line of the inter-specification page.

The sentence ``Figure 2C and the second diagram are'' is corrected to ``Figure 2C and 2D are''.

(5) From page 70, line 17 of the specification to page 71, line 3
In the line ``Figure 3C, Figure 3C'...(omitted)...
6. Correct the sentence 6 as follows:

3C(1), 3C(2), 30(1), 30(2), and 3E to 31(1) and 31(2). 3A (1) Standard P in the figure
The results of applying the filtering method according to the present invention to the TH formulation are shown in Fig. 3C(2) and Fig. 3D.
(2) and Figure 3I(2) are enlarged versions of the results shown in Figures 3C(1), 3D(1) and 3I(1);

Claims (6)

[Claims] (1) Separate, linear, transformation-constant filter function a^α
_N to obtain a baseline-corrected chromatogram, where N is a measure of the filter function width, α is a parameter whose value determines the signal-to-noise characteristics of the filter function, and the first chromatographic analysis of the sample to obtain a chromatogram of the first chromatogram, N being set to an approximation of the width of the peak obtained in the first chromatogram, and α being the second chromatogram with the first filtered baseline. said first chromatogram by a filter function set to filter out high frequency noise from said first chromatogram to obtain a chromatogram of
a third chromatogram having a baseline that is substantially the same as the first filtered baseline, with N set to an approximation of the peak width obtained in the first chromatogram;
filtering the second chromatogram by a filter function that sets α to the resolution increasing peak in the second chromatogram to obtain a chromatogram; A method comprising the step of subtracting said second chromatogram from a third chromatogram. (2) Truncating the fourth chromatogram for values below a preselected threshold to obtain a fifth chromatogram, setting α to remove high frequency noise, and obtaining a sixth chromatogram. filtering the fifth chromatogram by a filter function with N set equal to the width of the peak in the fifth chromatogram to obtain a peak in the sixth chromatogram; 2. The method of claim 1, comprising measuring the height of the peak. 3. The method of claim 2, wherein peaks detected in the sixth chromatogram are compared with peaks in a reference chromatogram to identify the sample. 4. The method of claim 3, wherein the sample comprises a mixture of amino acid derivatives. (5) The filter function is (N+1-|n|)2α/N(N+1)-{2α(N+
1)-N/N(2N+1)}. 5. A method according to claim 2, characterized in that the filter function is a shifted triangular filter function proportional to 1)-N/N(2N+1)}. (6) An apparatus for obtaining a baseline-corrected chromatogram of a sample having multiple components, comprising: separation means for separating and eluting the components at different times; and detecting the components to be eluted and separating the components. detection means for supplying a signal to said component corresponding to a retention time in said means (said signal hereinafter referred to as an unmodified chromatogram); and a computer coupled to said detection means for analyzing said unmodified chromatogram. means for filtering the unmodified chromatogram by a first filter function a^α_N that is a discrete, linear, transformation-constant function, where N is a measurement of the filter function width, set to an approximate value of the width of the peak obtained in said unmodified chromatogram, with filter means such that α is a parameter whose value determines the signal-to-noise characteristics of the filter function; to the unmodified chromatogram with N and α set to filter out high frequency noise from the unmodified chromatogram to obtain a second chromatogram with a first filtered baseline. control means for applying said filter means, said N being set to an approximation of the width of the peak obtained in said first chromatogram and substantially the same as said first filtered baseline; α was set to the resolution increasing peak in the second chromatogram to obtain a third chromatogram with a baseline of
and means for adding the filter means to the second chromatogram having a baseline-corrected fourth chromatogram; and control means also comprising means for subtracting from.